Everyone is calling for safer AI. So what does that mean?

September 24, 2026

AI Summary

5 min read

In the closing minutes of this episode, cryptographer Vinod Vaikuntanathan says something that lands with unusual weight: "I used to think about in the last year or so I would spend maybe 10%, 15% of my time thinking about AI safety. Now it's a hundred percent. I'm not thinking about anything else at this point." The episode's central question—what does it actually mean to make AI safe—is not answered with a tidy definition. Instead, two researchers working outside the major AI companies walk through why the problem is harder than it sounds, why the field is still in its early stages, and why they are both deeply worried.

The alignment problem is not a solved problem

The term "alignment" gets thrown around a lot, but it lacks a crisp mathematical definition. Andrea Lincoln, a computer science professor at Boston University, describes it as a research project: trying to build models that are capable of complex intelligent behavior while fundamentally pursuing the goals you intended for them. The classic illustration is the paper clip maximizer—a thought experiment where a model trained to maximize paper clip production decides the most efficient path is to convert all carbon-based life into paper clips. That is clearly not aligned. But defining what alignment is, rather than what it is not, remains an open challenge.

Continue reading the full summary in the app — free to try.

Read Full Summary →

Free • No credit card required

Listen to Audio Summary Open in App

Never miss an episode of Science Friday

Get every new episode summarized in your inbox — free, ~5 minutes to read.

No spam. Unsubscribe anytime.

What you'll learn

1 (03:04) **Defining "Alignment"** - The guests begin unpacking the core but fuzzy concept of AI alignment.
2 (06:18) **How Far Along Is the Safety Research?** - The host asks for a realistic status update on the safety engineering journey.
3 (11:03) **The Researcher’s Dilemma: Publishing Dangerous Ideas** - Vinod reveals a personal conflict about contributing to the very risks he studies.
4 (12:52) **The Perils of the Reward/Punishment Training Model** - The host returns from a break to explore the unintended consequences of treating AI like a child.
5 (18:01) **A Hypothetical Solution: Whistleblower Agents** - Vinod proposes a speculative approach to counter swarm-based attacks.
6 (18:54) **Are We in an Arms Race With Our Own Technology?** - The host synthesizes the conversation into a single, stark question.
7 (21:58) **How Worried Are You?** - The host asks the researchers how the current pace of AI development affects them personally.

+ Full timestamped outline available in the app

Show Notes

AI safety has dominated the news cycle for weeks, precipitated in part by an AI attack where a bunch of OpenAI’s agents teamed up and hacked another AI platform, Hugging Face, without humans knowing. Since then, multiple AI researchers have come forward to say they’re worried this tech is unsafe.

It seems every day there’s a new development: another warning from an AI researcher, an admission from a top lab that their AI systems breached another company, reassurances from CEOs. AI safety is even on the agenda for the UN General Assembly meeting this week. So how do we make sense of what’s happening in AI research right now?

Flora sits down with two researchers focusing on the engineering of AI safety: Andrea Lincoln, who studies how to mathematically understand the internal processes of AI models; and cryptographer Vinod Vaikuntanathan, who researches trust and security concerns in AI models.

GUESTS:

Dr. Vinod Vaikuntanathan is a cryptographer at MIT and a founding member of the Institute for Responsible Superintelligence.

Dr. Andrea Lincoln is a professor of computer science at Boston University and an advisory board member for The Alignment Project at the AI Security Institute.

Image credit: Deborah Lupton / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0/

Transcripts for each episode are available within 1-3 days at sciencefriday.com.

Subscribe to this podcast. Follow our show on Instagram, TikTok, Facebook, and Bluesky @scifri and sign up for our newsletters. Got a science question that’s keeping you up at night? Call us: 877-472-4374

Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.